Table of Contents
Every expression written in the Java programming language either produces no result (§15.1) or has a type that can be deduced at compile time (§15.3). When an expression appears in most contexts, it must be compatible with a type expected in that context; this type is called the target type. For convenience, compatibility of an expression with its surrounding context is facilitated in two ways:
First, for some expressions, termed poly expressions (§15.2), the deduced type can be influenced by the target type. The same expression can have different types in different contexts.
Second, after the type of the expression has been deduced, an implicit conversion from the type of the expression to the target type can sometimes be performed.
If neither strategy is able to produce the appropriate type, a compile-time error occurs.
The rules determining whether an expression is a poly expression, and if so, its type and compatibility in a particular context, vary depending on the kind of context and the form of the expression. In addition to influencing the type of the expression, the target type may in some cases influence the run time behavior of the expression in order to produce a value of the appropriate type.
Similarly, the rules determining whether a target type allows an implicit conversion vary depending on the kind of context, the type of the expression, and, in one special case, the value of a constant expression (§15.29). A conversion from type S to type T allows an expression of type S to be treated at compile time as if it had type T instead. In some cases this will require a corresponding action at run time to check the validity of the conversion or to translate the run-time value of the expression into a form appropriate for the new type T.
Example 5.0-1. Conversions at Compile Time and Run Time
A conversion from type Object
to type Thread
requires a run-time check to make sure that the run-time value
is actually an instance of class Thread
or one of its
subclasses; if it is not, an exception is thrown.
A conversion from type Thread
to type Object
requires no run-time action; Thread
is a subclass of Object
,
so any reference produced by an expression of type Thread
is a
valid reference value of type Object
.
A conversion from type int
to type long
requires run-time sign-extension of a 32-bit integer value to
the 64-bit long
representation. No information is lost.
A conversion from type double
to type long
requires a non-trivial translation from a 64-bit floating-point
value to the 64-bit integer representation. Depending on the
actual run-time value, information may be lost.
The conversions possible in the Java programming language are grouped into several broad categories:
There are six kinds of conversion contexts in which poly expressions may be influenced by context or implicit conversions may occur. Each kind of context has different rules for poly expression typing and allows conversions in some of the categories above but not others. The contexts are:
Assignment contexts (§5.2, §15.26), in which an expression's value is bound to a named variable. Primitive and reference types are subject to widening, values may be boxed or unboxed, and some primitive constant expressions may be subject to narrowing. An unchecked conversion may also occur.
Strict invocation contexts (§5.3, §15.9, §15.12), in which an argument is bound to a formal parameter of a constructor or method. Widening primitive, widening reference, and unchecked conversions may occur.
Loose invocation contexts (§5.3, §15.9, §15.12), in which, like strict invocation contexts, an argument is bound to a formal parameter. Method or constructor invocations may provide this context if no applicable declaration can be found using only strict invocation contexts. In addition to widening and unchecked conversions, this context allows boxing and unboxing conversions to occur.
String contexts (§5.4,
§15.18.1), in which a value of any type is
converted to an object of type String
.
Casting contexts (§5.5), in which an expression's value is converted to a type explicitly specified by a cast operator (§15.16). Casting contexts are more inclusive than assignment or loose invocation contexts, allowing any specific conversion other than a string conversion, but certain casts to a reference type are checked for correctness at run time.
Numeric contexts (§5.6), in which the operands of a numeric operator or some other expressions that operate on numbers may be widened to a common type.
The term "conversion" is also used to describe, without being specific, any conversions allowed in a particular context. For example, we say that an expression that is the initializer of a local variable is subject to "assignment conversion", meaning that a specific conversion will be implicitly chosen for that expression according to the rules for the assignment context. As another example, we say that an expression undergoes "casting conversion" to mean that the expression's type will be converted as permitted in a casting context.
Example 5.0-2. Conversions In Various Contexts
class Test { public static void main(String[] args) { // Casting conversion (5.5) of a float literal to // type int. Without the cast operator, this would // be a compile-time error, because this is a // narrowing conversion (5.1.3): int i = (int)12.5f; // String conversion (5.4) of i's int value: System.out.println("(int)12.5f==" + i); // Assignment conversion (5.2) of i's value to type // float. This is a widening conversion (5.1.2): float f = i; // String conversion of f's float value: System.out.println("after float widening: " + f); // Numeric promotion (5.6) of i's value to type // float. This is a binary numeric promotion. // After promotion, the operation is float*float: System.out.print(f); f = f * i; // Two string conversions of i and f: System.out.println("*" + i + "==" + f); // Invocation conversion (5.3) of f's value // to type double, needed because the method Math.sin // accepts only a double argument: double d = Math.sin(f); // Two string conversions of f and d: System.out.println("Math.sin(" + f + ")==" + d); } }
This program produces the output:
(int)12.5f==12 after float widening: 12.0 12.0*12==144.0 Math.sin(144.0)==-0.49102159389846934
Specific type conversions in the Java programming language are divided into 12 kinds.
A conversion from a type to that same type is permitted for any type.
This may seem trivial, but it has two practical consequences. First, it is always permitted for an expression to have the desired type to begin with, thus allowing the simply stated rule that every expression is subject to conversion, if only a trivial identity conversion. Second, it implies that it is permitted for a program to include redundant cast operators for the sake of clarity.
19 specific conversions on primitive types are called the widening primitive conversions:
A widening primitive conversion does not lose information about the overall magnitude of a numeric value in the following cases, where the numeric value is preserved exactly:
A widening primitive conversion from int
to float
, or from long
to
float
, or from long
to double
, may result in loss of
precision, that is, the result may lose some of the least
significant bits of the value. In this case, the resulting
floating-point value will be a correctly rounded version of the
integer value, using
the round to nearest rounding policy (§15.4).
A widening conversion of a signed integer value to an integral type T simply sign-extends the two's-complement representation of the integer value to fill the wider format.
A widening conversion of a char
to an integral type T zero-extends
the representation of the char
value to fill the wider
format.
A widening conversion from int
to float
, or from long
to float
,
or from int
to double
, or from long
to double
occurs as
determined by the rules of IEEE 754 for converting from an
integer format to a binary floating-point format.
A widening conversion from float
to double
occurs as determined by
the rules of IEEE 754 for converting between binary floating-point
formats.
Despite the fact that loss of precision may occur, a widening primitive conversion never results in a run-time exception (§11.1.1).
Example 5.1.2-1. Widening Primitive Conversion
class Test { public static void main(String[] args) { int big = 1234567890; float approx = big; System.out.println(big - (int)approx); } }
This program prints:
-46
thus indicating that information was lost during the
conversion from type int
to type float
because values of type
float
are not precise to nine significant digits.
22 specific conversions on primitive types are called the narrowing primitive conversions:
A narrowing primitive conversion may lose information about the overall magnitude of a numeric value, and may also lose precision and range.
A narrowing conversion of a signed integer to an integral type T simply discards all but the n lowest order bits, where n is the number of bits used to represent type T. In addition to a possible loss of information about the magnitude of the numeric value, this may cause the sign of the resulting value to differ from the sign of the input value.
A narrowing conversion of a char
to an integral type T likewise
simply discards all but the n lowest order bits, where n is
the number of bits used to represent type T. In addition to a
possible loss of information about the magnitude of the numeric value,
this may cause the resulting value to be a negative number, even
though chars represent 16-bit unsigned integer values.
A narrowing conversion of a floating-point number to an integral type T takes two steps:
In the first step, the floating-point number is converted either
to a long
, if T is long
, or to an int
, if T is byte
,
short
, char
, or int
, as follows:
If the floating-point number is NaN (§4.2.3),
the result of the first step of the conversion is an int
or
long
0
.
Otherwise, if the floating-point number is not an infinity,
the floating-point value is rounded to an integer value
V
using the round toward
zero rounding policy (§4.2.4).
Then there are two cases:
Otherwise, one of the following two cases must be true:
The value must be too small
(a negative value of large magnitude or negative
infinity), and the result of the first step is the
smallest representable value of type int
or
long
.
The value must be too large
(a positive value of large magnitude or positive
infinity), and the result of the first step is the
largest representable value of type int
or
long
.
If T is int
or long
, the result of the conversion
is the result of the first step.
If T is byte
, char
, or short
, the result of the
conversion is the result of a narrowing conversion to type T
(§5.1.3) of the result of the first
step.
A narrowing conversion from double
to float
occurs as determined by the rules of IEEE 754 for converting between
binary floating-point formats, using the round to nearest rounding
policy (§15.4). This conversion can lose
precision, but also lose range, resulting in a float
zero from a
nonzero double
and a float
infinity from a finite double
. A
double
NaN is converted to a float
NaN and a double
infinity is
converted to the same-signed float
infinity.
Despite the fact that overflow, underflow, or other loss of information may occur, a narrowing primitive conversion never results in a run-time exception (§11.1.1).
Example 5.1.3-1. Narrowing Primitive Conversion
class Test { public static void main(String[] args) { float fmin = Float.NEGATIVE_INFINITY; float fmax = Float.POSITIVE_INFINITY; System.out.println("long: " + (long)fmin + ".." + (long)fmax); System.out.println("int: " + (int)fmin + ".." + (int)fmax); System.out.println("short: " + (short)fmin + ".." + (short)fmax); System.out.println("char: " + (int)(char)fmin + ".." + (int)(char)fmax); System.out.println("byte: " + (byte)fmin + ".." + (byte)fmax); } }
This program produces the output:
long: -9223372036854775808..9223372036854775807 int: -2147483648..2147483647 short: 0..-1 char: 0..65535 byte: 0..-1
The results for char
, int
, and long
are
unsurprising, producing the minimum and maximum representable values
of the type.
The results for byte
and short
lose information
about the sign and magnitude of the numeric values and also lose
precision. The results can be understood by examining the low order
bits of the minimum and maximum int
. The minimum int
is, in
hexadecimal, 0x80000000
, and the maximum int
is 0x7fffffff
. This explains the short
results,
which are the low 16 bits of these values,
namely, 0x0000
and 0xffff
; it
explains the char results, which also are the low 16 bits of these
values, namely, '\u0000'
and '\uffff'
; and it explains the byte results,
which are the low 8 bits of these values,
namely, 0x00
and 0xff
.
Example 5.1.3-2. Narrowing Primitive Conversions that lose information
class Test { public static void main(String[] args) { // A narrowing of int to short loses high bits: System.out.println("(short)0x12345678==0x" + Integer.toHexString((short)0x12345678)); // An int value too big for byte changes sign and magnitude: System.out.println("(byte)255==" + (byte)255); // A float value too big to fit gives largest int value: System.out.println("(int)1e20f==" + (int)1e20f); // A NaN converted to int yields zero: System.out.println("(int)NaN==" + (int)Float.NaN); // A double value too large for float yields infinity: System.out.println("(float)-1e100==" + (float)-1e100); // A double value too small for float underflows to zero: System.out.println("(float)1e-50==" + (float)1e-50); } }
This program produces the output:
(short)0x12345678==0x5678 (byte)255==-1 (int)1e20f==2147483647 (int)NaN==0 (float)-1e100==-Infinity (float)1e-50==0.0
The following conversion combines both widening and narrowing primitive conversions:
First,
the byte
is converted to an int
via widening primitive conversion
(§5.1.2), and then the resulting int
is
converted to a char
by narrowing primitive conversion
(§5.1.3).
A widening reference conversion exists from any reference type S to any reference type T, provided S is a subtype of T (§4.10).
Widening reference conversions never require a special action at run time and therefore never throw an exception at run time. They consist simply in regarding a reference as having some other type in a manner that can be proved correct at compile time.
The null type is not a reference type (§4.1), and so a widening reference conversion does not exist from the null type to a reference type. However, many conversion contexts explicitly allow the null type to be converted to a reference type.
A narrowing reference conversion treats expressions of a reference type S as expressions of a different reference type T, where S is not a subtype of T. The supported pairs of types are defined in §5.1.6.1. Unlike widening reference conversion, the types need not be directly related. However, there are restrictions that prohibit conversion between certain pairs of types when it can be statically proven that no value can be of both types.
A narrowing reference conversion may require a test at run time to
validate that a value of type S is a legitimate value of type
T. However, due to the lack of parameterized type information at run
time, some conversions cannot be fully validated by a run time test;
they are flagged at compile time (§5.1.6.2). For
conversions that can be fully validated by a run time test, and for
certain conversions that involve parameterized type information but
can still be partially validated at run time, a ClassCastException
is thrown if the
test fails (§5.1.6.3).
A narrowing reference conversion exists from reference type S to reference type T if all of the following are true:
S is not a subtype of T (§4.10)
If there exists a parameterized type X that is a supertype of T, and a parameterized type Y that is a supertype of S, such that the erasures of X and Y are the same, then X and Y are not provably distinct (§4.5).
Using types from the java.util
package as an
example, no narrowing reference conversion exists from
ArrayList<String>
to ArrayList<Object>
, or vice versa,
because the type arguments String
and Object
are provably
distinct. For the same reason, no narrowing reference conversion
exists from ArrayList<String>
to List<Object>
, or vice versa. The
rejection of provably distinct types is a simple static gate to
prevent "stupid" narrowing reference conversions.
One of the following cases applies:
S is a class or interface type, and T is a class or interface type, and S names a class or interface that is not disjoint from the class or interface named by T. ("disjoint" is defined below.)
S is the class type Object
or the interface type
java.io.Serializable
or Cloneable
(the only interfaces
implemented by arrays (§10.8)), and T
is an array type.
S is an array type SC[]
, that is, an array of
components of type SC; T is an array type TC[]
,
that is, an array of components of type TC; and a narrowing
reference conversion exists from SC to TC.
S is a type variable, and a narrowing reference conversion exists from the upper bound of S to T.
T is a type variable, and either a widening reference conversion or a narrowing reference conversion exists from S to the upper bound of T.
S is an intersection type S1 &
... &
Sn, and
for all i (1 ≤ i ≤ n), either a
widening reference conversion or a narrowing reference
conversion exists from Si to T.
T is an intersection type T1 &
... &
Tn, and
for all i (1 ≤ i ≤ n), either a
widening reference conversion or a narrowing reference
conversion exists from S to Ti.
A class or interface is disjoint from another
class or interface if it can be determined statically that they have
no instances in common (other than the null
value). The rules for
disjointess are as follows:
A class named C is disjoint from an interface named I
if (i) it is not the case that C <:
I, and (ii) one of the
following cases applies:
C is sealed
, and all of the permitted direct subclasses of
C are disjoint from I.
C is freely extensible (§8.1.1.2),
and I is sealed
, and C is disjoint from all of the
permitted direct subclasses and subinterfaces of I.
An interface named I is disjoint from a class named C if C is disjoint from I.
A class named C is disjoint from another class named D if
(i) it is not the case that C <:
D, and
(ii) it is not the case that D <:
C.
An interface named I is disjoint from another interface named J if
(i) it is not that case that I <:
J, and
(ii) it is not the case that J <:
I, and
(iii) one of the following cases applies:
Whether a class is final
has the most bearing on
whether the class is disjoint from interfaces. Consider the following
declarations:
interface I {} final class C {}
As class C is final
and does not implement I,
there can be no instances of C that are also an instance of I, so
C and I are disjoint. Therefore, there is no narrowing reference
conversion from C to I.
In contrast, consider the following declarations:
interface J {} class D {}
Even though class D does not implement J, it is still possible for an instance of D to be an instance of J, for example, if the following declaration occurs:
class E extends D implements J {}
For this reason, D is not disjoint from J, and there is a narrowing reference conversion from D to J.
The final clause above implies that two freely extensible interfaces (§9.1.1.4) are not disjoint.
A narrowing reference conversion is either checked or unchecked. These terms refer to the ability of the Java Virtual Machine to validate, or not, the type correctness of the conversion.
If a narrowing reference conversion is unchecked, then the Java Virtual Machine will
not be able to fully validate its type correctness, possibly leading
to heap pollution (§4.12.2). To flag this to the
programmer, an unchecked narrowing reference conversion causes a
compile-time unchecked warning, unless suppressed
by @SuppressWarnings
(§9.6.4.5). Conversely,
if a narrowing reference conversion is not unchecked, then it is
checked; the Java Virtual Machine will be able to fully validate its type
correctness, so no warning is given at compile time.
The unchecked narrowing reference conversions are as follows:
A narrowing reference conversion from a type S to a parameterized class or interface type T is unchecked, unless at least one of the following is true:
A narrowing reference conversion from a type S to a type variable T is unchecked.
A narrowing reference conversion from a type S to an
intersection type T1 &
... &
Tn is unchecked if
there exists a Ti (1 ≤ i ≤ n) such that S
is not a subtype of Ti and a narrowing reference conversion
from S to Ti is unchecked.
All checked narrowing reference conversions require a validity check at run time. Primarily, these conversions are to class and interface types that are not parameterized.
Some unchecked narrowing reference conversions require a validity check at run time. This depends on whether the unchecked narrowing reference conversion is completely unchecked or partially unchecked. A partially unchecked narrowing reference conversion requires a validity check at run time, while a completely unchecked narrowing reference conversion does not.
These terms refer to the compatibility of the types involved in the conversion when viewed as raw types. If the conversion is conceptually an "upcast", then the conversion is completely unchecked; no run time test is needed because the conversion is legal in the non-generic type system of the Java Virtual Machine. Conversely, if the conversion is conceptually a "downcast", then the conversion is partially unchecked; even in the non-generic type system of the Java Virtual Machine, a run time check is needed to test the compatibility of the (raw) types involved in the conversion.
Using types from the java.util
package as an
example, a conversion from ArrayList<String>
to Collection<T>
is completely unchecked,
because the (raw) type ArrayList
is a subtype of
the (raw) type Collection
in the Java Virtual Machine. Conversely,
a conversion from Collection<T>
to ArrayList<String>
is partially unchecked,
because the (raw) type Collection
is not a subtype
of the (raw) type ArrayList
in the Java Virtual Machine.
The categorization of an unchecked narrowing reference conversion is as follows:
An unchecked narrowing reference conversion from S to a
non-intersection type T is completely unchecked if
|S| <:
|T|.
An unchecked narrowing reference conversion from S to an
intersection type T1 &
... &
Tn is completely
unchecked if, for all i (1 ≤ i ≤ n),
either S <:
Ti or a narrowing reference conversion
from S to Ti is completely unchecked.
The run time validity check for a checked or partially unchecked narrowing reference conversion is as follows:
If the value at run time is null
, then the conversion is
allowed.
Otherwise, let R be the class of the object referred to by the value, and let T be the erasure (§4.6) of the type being converted to. Then:
Note that R cannot be an interface when these rules are first applied for any given conversion, but R may be an interface if the rules are applied recursively because the run-time reference value may refer to an array whose element type is an interface type.
If T is a class type, then T must be Object
(§4.3.2), or a ClassCastException
is thrown.
If T is an interface type, then R must be either the same
interface as T or a subinterface of T, or a ClassCastException
is thrown.
If R is a class representing an array type RC[]
,
that is, an array of components of type RC:
If T is a class type, then T must be Object
(§4.3.2), or a ClassCastException
is thrown.
If T is an interface type, then T must be the type
java.io.Serializable
or Cloneable
(the only interfaces
implemented by arrays), or a ClassCastException
is thrown.
If T is an array type TC[]
, that is, an
array of components of type TC, then a ClassCastException
is thrown
unless either TC and RC are the same primitive type,
or TC and RC are reference types and are allowed by
a recursive application of these run-time rules.
If the conversion is to an intersection type T1 &
... &
Tn, then for all i (1 ≤ i ≤ n), any run-time
check required for a conversion from S to Ti is also required for
the conversion to the intersection type.
Boxing conversion treats expressions of a primitive type as expressions of a corresponding reference type. Specifically, the following nine conversions are called the boxing conversions:
From the null type to the null type
This rule is necessary because the conditional operator (§15.25) applies boxing conversion to the types of its operands, and uses the result in further calculations.
At run time, boxing conversion proceeds as follows:
If p
is a value of type boolean
, then boxing conversion
converts p
into a reference r
of class and type Boolean
,
such that r
.booleanValue() == p
If p
is a value of type byte
, then boxing conversion
converts p
into a reference r
of class and type Byte
, such
that r
.byteValue() == p
If p
is a value of type char
, then boxing conversion
converts p
into a reference r
of class and type Character
,
such that r
.charValue() == p
If p
is a value of type short
, then boxing conversion
converts p
into a reference r
of class and type Short
,
such that r
.shortValue() == p
If p
is a value of type int
, then boxing conversion converts
p
into a reference r
of class and type Integer
, such
that r
.intValue() == p
If p
is a value of type long
, then boxing conversion
converts p
into a reference r
of class and type Long
, such
that r
.longValue() == p
If p
is a value of any other type, boxing conversion is
equivalent to an identity conversion
(§5.1.1).
If the value p
being boxed is the result of evaluating a constant
expression (§15.29) of type boolean
, byte
, char
,
short
, int
, or long
, and the result is true
, false
, a
character in the range '\u0000'
to '\u007f'
inclusive, or an integer in the
range -128
to 127
inclusive,
then let a
and b
be the results of any two boxing conversions of
p
. It is always the case that a
==
b
.
Ideally, boxing a primitive value would always yield an identical reference. In practice, this may not be feasible using existing implementation techniques. The rule above is a pragmatic compromise, requiring that certain common values always be boxed into indistinguishable objects. The implementation may cache these, lazily or eagerly. For other values, the rule disallows any assumptions about the identity of the boxed values on the programmer's part. This allows (but does not require) sharing of some or all of these references.
This ensures that in most common cases, the behavior
will be the desired one, without imposing an undue performance
penalty, especially on small devices. Less memory-limited
implementations might, for example, cache all char
and short
values, as well as int
and long
values in the range of -32K to
+32K.
A boxing
conversion may result in an OutOfMemoryError
if a new instance of one of the
wrapper classes (Boolean
, Byte
, Character
, Short
, Integer
,
Long
, Float
, or Double
) needs to be allocated and insufficient
storage is available.
Unboxing conversion treats expressions of a reference type as expressions of a corresponding primitive type. Specifically, the following eight conversions are called the unboxing conversions:
At run time, unboxing conversion proceeds as follows:
If r
is a reference of type Boolean
, then unboxing
conversion converts r
into r
.booleanValue()
If r
is a reference of type Byte
, then unboxing conversion
converts r
into r
.byteValue()
If r
is a reference of type Character
, then unboxing
conversion converts r
into r
.charValue()
If r
is a reference of type Short
, then unboxing conversion
converts r
into r
.shortValue()
If r
is a reference of type Integer
, then unboxing
conversion converts r
into r
.intValue()
If r
is a reference of type Long
, then unboxing conversion
converts r
into r
.longValue()
If r
is a reference of type Float
, unboxing conversion
converts r
into r
.floatValue()
If r
is a reference of type Double
, then unboxing conversion
converts r
into r
.doubleValue()
If r
is null
, unboxing conversion throws a NullPointerException
A type is said to be convertible to a numeric type if it is a numeric type (§4.2), or it is a reference type that may be converted to a numeric type by unboxing conversion.
A type is said to be convertible to an integral type if it is an integral type, or it is a reference type that may be converted to an integral type by unboxing conversion.
Let G name a generic type declaration with n type parameters.
There is an unchecked conversion from the
raw class or interface type (§4.8)
G to any parameterized type of the form G<
T1,...,Tn>
.
There is an unchecked conversion from the raw
array type G[]
k to any array type
of the form G<
T1,...,Tn>
[]
k.
(The notation []
k indicates an
array type of k dimensions.)
Use of an unchecked conversion causes a
compile-time unchecked warning unless
all type arguments Ti (1 ≤ i ≤ n) are unbounded
wildcards (§4.5.1), or the warning is suppressed
by @SuppressWarnings
(§9.6.4.5).
Unchecked conversion is used to enable a smooth
interoperation of legacy code, written before the introduction of
generic types, with libraries that have undergone a conversion to use
genericity (a process we call generification). In such circumstances
(most notably, clients of the Collections Framework
in java.util
), legacy code uses raw types
(e.g. Collection
instead
of Collection<String>
). Expressions of raw
types are passed as arguments to library methods that use
parameterized versions of those same types as the types of their
corresponding formal parameters.
Such calls cannot be shown to be statically safe under the type system using generics. Rejecting such calls would invalidate large bodies of existing code, and prevent them from using newer versions of the libraries. This in turn, would discourage library vendors from taking advantage of genericity. To prevent such an unwelcome turn of events, a raw type may be converted to an arbitrary invocation of the generic type declaration to which the raw type refers. While the conversion is unsound, it is tolerated as a concession to practicality. An unchecked warning is issued in such cases.
Let G name a generic type declaration (§8.1.2, §9.1.2) with n type parameters A1,...,An with corresponding bounds U1,...,Un.
There exists a capture conversion
from a parameterized type G<
T1,...,Tn>
(§4.5)
to a parameterized type G<
S1,...,Sn>
,
where, for 1 ≤ i ≤ n :
If Ti is a wildcard type argument (§4.5.1)
of the form ?
, then Si is a fresh type variable
whose upper bound is Ui[A1:=S1,...,An:=Sn]
and whose lower bound is the null type (§4.1).
If Ti is a wildcard type argument of the form ?
extends
Bi, then Si is a fresh type variable whose upper bound is
glb(Bi, Ui[A1:=S1,...,An:=Sn]
) and
whose lower bound is the null type.
glb(V1,...,Vm) is defined as V1 &
... &
Vm.
It is a compile-time error if, for any two classes (not interfaces) Vi and Vj, Vi is not a subclass of Vj or vice versa.
If Ti is a wildcard type argument of the form ?
super
Bi, then Si is a fresh type variable whose upper bound is
Ui[A1:=S1,...,An:=Sn]
and
whose lower bound is Bi.
Capture conversion on any type other than a parameterized type (§4.5) acts as an identity conversion (§5.1.1).
Capture conversion is not applied recursively.
Capture conversion never requires a special action at run time and therefore never throws an exception at run time.
Capture conversion is designed to make wildcards
more useful. To understand the motivation, let's begin by looking at
the method java.util.Collections.reverse()
:
public static void reverse(List<?> list);
The method reverses the list provided as a
parameter. It works for any type of list, and so the use of the
wildcard type List<?>
as the type of the
formal parameter is entirely appropriate.
Now consider how one would
implement reverse()
:
public static void reverse(List<?> list) { rev(list); } private static <T> void rev(List<T> list) { List<T> tmp = new ArrayList<T>(list); for (int i = 0; i < list.size(); i++) { list.set(i, tmp.get(list.size() - i - 1)); } }
The implementation needs to copy the list, extract
elements from the copy, and insert them into the original. To do this
in a type-safe manner, we need to give a name, T
,
to the element type of the incoming list. We do this in the private
service method rev()
. This requires us to pass the
incoming argument list, of type List<?>
, as
an argument to rev()
. In
general, List<?>
is a list of unknown
type. It is not a subtype of List<T>
, for any
type T. Allowing such a subtype relation would be unsound. Given the
method:
public static <T> void fill(List<T> l, T obj)
the following code would undermine the type system:
List<String> ls = new ArrayList<String>(); List<?> l = ls; Collections.fill(l, new Object()); // not legal - but assume it was! String s = ls.get(0); // ClassCastException - ls contains // Objects, not Strings.
So, without some special dispensation, we can see
that the call from reverse()
to rev()
would be disallowed. If this were the
case, the author of reverse()
would be forced to
write its signature as:
public static <T> void reverse(List<T> list)
This is undesirable, as it exposes implementation information to the caller. Worse, the designer of an API might reason that the signature using a wildcard is what the callers of the API require, and only later realize that a type safe implementation was precluded.
The call from reverse()
to rev()
is in fact harmless, but it cannot be
justified on the basis of a general subtyping relation
between List<?>
and List<T>
. The call is harmless, because
the incoming argument is doubtless a list of some type (albeit an
unknown one). If we can capture this unknown type in a type
variable X
, we can infer T
to
be X
. That is the essence of capture
conversion. The specification of course must cope with complications,
like non-trivial (and possibly recursively defined) upper or lower
bounds, the presence of multiple arguments etc.
Mathematically sophisticated readers will want to relate capture conversion to established type theory. Readers unfamiliar with type theory can skip this discussion - or else study a suitable text, such as Types and Programming Languages by Benjamin Pierce, and then revisit this section.
Here then is a brief summary of the relationship of
capture conversion to established type theoretical notions. Wildcard
types are a restricted form of existential types. Capture conversion
corresponds loosely to an opening of a value of existential type. A
capture conversion of an expression e
can be
thought of as an open
of e
in a
scope that comprises the top level expression that
encloses e
.
The classical open
operation on
existentials requires that the captured type variable must not escape
the opened expression. The open
that corresponds to
capture conversion is always on a scope sufficiently large that the
captured type variable can never be visible outside that scope. The
advantage of this scheme is that there is no need for
a close
operation, as defined in the
paper On Variance-Based Subtyping for Parametric
Types by Atsushi Igarashi and Mirko Viroli, in the
proceedings of the 16th European Conference on Object Oriented
Programming (ECOOP 2002). For a formal account of wildcards,
see Wild FJ by Mads Torgersen, Erik Ernst and
Christian Plesner Hansen, in the 12th workshop on Foundations of
Object Oriented Programming (FOOL 2005).
Any type may be converted to type String
by string
conversion.
A value x
of primitive type T is first converted
to a reference value as if by giving it as an argument to an
appropriate class instance creation expression (§15.9):
This reference value is then converted to type String
by string
conversion.
Now only reference values need to be considered:
If the reference is null
, it is converted to the string
"null
" (four ASCII characters n
,
u
, l
, l
).
Otherwise, the conversion is performed as if by an invocation of
the toString
method of the referenced object with no
arguments; but if the result of invoking the toString
method
is null
, then the string "null
" is used instead.
The toString
method is defined by the
primordial class Object
(§4.3.2). Many
classes override it, notably Boolean
, Character
, Integer
,
Long
, Float
, Double
, and String
.
Assignment contexts allow the value of an expression to be assigned (§15.26) to a variable; the type of the expression must be converted to the type of the variable.
Assignment contexts allow the use of one of the following:
an identity conversion (§5.1.1)
a widening primitive conversion (§5.1.2)
a widening reference conversion (§5.1.5)
a widening reference conversion followed by an unboxing conversion
a widening reference conversion followed by an unboxing conversion, then followed by a widening primitive conversion
a boxing conversion (§5.1.7)
a boxing conversion followed by a widening reference conversion
an unboxing conversion (§5.1.8)
an unboxing conversion followed by a widening primitive conversion
If, after the conversions listed above have been applied, the resulting type is a raw type (§4.8), an unchecked conversion (§5.1.9) may then be applied.
In addition, if the expression is a constant expression
(§15.29) of type byte
, short
, char
, or int
:
A narrowing primitive conversion may be used if the variable is
of type byte
, short
, or char
, and the value of the
constant expression is representable in the type of the variable.
A narrowing primitive conversion followed by a boxing conversion
may be used if the variable is of type Byte
, Short
, or
Character
, and the value of the constant expression is
representable in the type byte
, short
, or char
respectively.
The compile-time narrowing of constant expressions means that code such as:
byte theAnswer = 42;
is allowed. Without the narrowing, the fact that the
integer literal 42
has type int
would mean that a
cast to byte
would be required:
byte theAnswer = (byte)42; // cast is permitted but not required
Finally, a value of the null type (the null reference is the only such value) may be assigned to any reference type, resulting in a null reference of that type.
It is a compile-time error if the chain of conversions contains two parameterized types that are not in the subtype relation (§4.10).
An example of such an illegal chain would be:
Integer, Comparable<Integer>, Comparable, Comparable<String>
The first three elements of the chain are related by
widening reference conversion, while the last entry is derived from
its predecessor by unchecked conversion. However, this is not a valid
assignment conversion, because the chain contains two parameterized
types, Comparable<Integer>
and Comparable<String>
, that are not
subtypes.
If the type of an expression can be converted to the type of a variable by assignment conversion, we say the expression (or its value) is assignable to the variable or, equivalently, that the type of the expression is assignment compatible with the type of the variable.
The only exceptions that may arise from conversions in an assignment context are:
A ClassCastException
if, after the conversions above have been applied, the
resulting value is an object which is not an instance of a
subclass or subinterface of the erasure
(§4.6) of the type of the variable.
This circumstance can only arise as a result of heap pollution (§4.12.2). In practice, implementations need only perform casts when accessing a field or method of an object of parameterized type when the erased type of the field, or the erased return type of the method, differ from its unerased type.
A NullPointerException
as a result of an unboxing conversion on a null
reference.
An ArrayStoreException
in special cases involving array elements or field
access (§10.5, §15.26.1).
Example 5.2-1. Assignment for Primitive Types
class Test { public static void main(String[] args) { short s = 12; // narrow 12 to short float f = s; // widen short to float System.out.println("f=" + f); char c = '\u0123'; long l = c; // widen char to long System.out.println("l=0x" + Long.toString(l,16)); f = 1.23f; double d = f; // widen float to double System.out.println("d=" + d); } }
This program produces the output:
f=12.0 l=0x123 d=1.2300000190734863
The following program, however, produces compile-time errors:
class Test { public static void main(String[] args) { short s = 123; char c = s; // error: would require cast s = c; // error: would require cast } }
because not all short
values are char
values,
and neither are all char
values short
values.
Example 5.2-2. Assignment for Reference Types
class Point { int x, y; } class Point3D extends Point { int z; } interface Colorable { void setColor(int color); } class ColoredPoint extends Point implements Colorable { int color; public void setColor(int color) { this.color = color; } } class Test { public static void main(String[] args) { // Assignments to variables of class type: Point p = new Point(); p = new Point3D(); // OK because Point3D is a subclass of Point Point3D p3d = p; // Error: will require a cast because a Point // might not be a Point3D (even though it is, // dynamically, in this example.) // Assignments to variables of type Object: Object o = p; // OK: any object to Object int[] a = new int[3]; Object o2 = a; // OK: an array to Object // Assignments to variables of interface type: ColoredPoint cp = new ColoredPoint(); Colorable c = cp; // OK: ColoredPoint implements Colorable // Assignments to variables of array type: byte[] b = new byte[4]; a = b; // Error: these are not arrays of the same primitive type Point3D[] p3da = new Point3D[3]; Point[] pa = p3da; // OK: since we can assign a Point3D to a Point p3da = pa; // Error: (cast needed) since a Point // can't be assigned to a Point3D } }
The following test program illustrates assignment conversions on reference values, but fails to compile, as described in its comments. This example should be compared to the preceding one.
class Point { int x, y; } interface Colorable { void setColor(int color); } class ColoredPoint extends Point implements Colorable { int color; public void setColor(int color) { this.color = color; } } class Test { public static void main(String[] args) { Point p = new Point(); ColoredPoint cp = new ColoredPoint(); // Okay because ColoredPoint is a subclass of Point: p = cp; // Okay because ColoredPoint implements Colorable: Colorable c = cp; // The following cause compile-time errors because // we cannot be sure they will succeed, depending on // the run-time type of p; a run-time check will be // necessary for the needed narrowing conversion and // must be indicated by including a cast: cp = p; // p might be neither a ColoredPoint // nor a subclass of ColoredPoint c = p; // p might not implement Colorable } }
Example 5.2-3. Assignment for Array Types
class Point { int x, y; } class ColoredPoint extends Point { int color; } class Test { public static void main(String[] args) { long[] veclong = new long[100]; Object o = veclong; // okay Long l = veclong; // compile-time error short[] vecshort = veclong; // compile-time error Point[] pvec = new Point[100]; ColoredPoint[] cpvec = new ColoredPoint[100]; pvec = cpvec; // okay pvec[0] = new Point(); // okay at compile time, // but would throw an // exception at run time cpvec = pvec; // compile-time error } }
In this example:
The value of veclong
cannot
be assigned to a Long
variable, because Long
is a class type
other than Object
. An array can be assigned only to a variable
of a compatible array type, or to a variable of type Object
,
Cloneable
or java.io.Serializable
.
The value of veclong
cannot
be assigned to vecshort
, because they are
arrays of primitive type, and short
and long
are not the
same primitive type.
The value of cpvec
can be
assigned to pvec
, because any reference that
could be the value of an expression of
type ColoredPoint
can be the value of a
variable of type Point
. The subsequent
assignment of the new Point
to a component
of pvec
then would throw an ArrayStoreException
(if the
program were otherwise corrected so that it could be compiled),
because a ColoredPoint
array cannot have an
instance of Point
as the value of a
component.
The value of pvec
cannot be
assigned to cpvec
, because not every
reference that could be the value of an expression of
type Point
can correctly be the value of a
variable of type ColoredPoint
. If the value
of pvec
at run time were a reference to an
instance of Point[]
, and the assignment
to cpvec
were allowed, a simple reference to
a component of cpvec
,
say, cpvec[0]
, could return
a Point
, and a Point
is
not a ColoredPoint
. Thus to allow such an
assignment would allow a violation of the type system. A cast
may be used (§5.5,
§15.16) to ensure
that pvec
references
a ColoredPoint[]
:
cpvec = (ColoredPoint[])pvec; // OK, but may throw an // exception at run time
Invocation contexts allow an argument value in a method or constructor invocation (§8.8.7.1, §15.9, §15.12) to be assigned to a corresponding formal parameter.
Strict invocation contexts allow the use of one of the following:
Loose invocation contexts allow a more permissive set of conversions, because they are only used for a particular invocation if no applicable declaration can be found using strict invocation contexts. Loose invocation contexts allow the use of one of the following:
an identity conversion (§5.1.1)
a widening primitive conversion (§5.1.2)
a widening reference conversion (§5.1.5)
a widening reference conversion followed by an unboxing conversion
a widening reference conversion followed by an unboxing conversion, then followed by a widening primitive conversion
a boxing conversion (§5.1.7)
a boxing conversion followed by widening reference conversion
an unboxing conversion (§5.1.8)
an unboxing conversion followed by a widening primitive conversion
If, after the conversions listed for an invocation context have been applied, the resulting type is a raw type (§4.8), an unchecked conversion (§5.1.9) may then be applied.
A value of the null type (the null reference is the only such value) may be assigned to any reference type.
It is a compile-time error if the chain of conversions contains two parameterized types that are not in the subtype relation (§4.10).
The only exceptions that may arise in an invocation context are:
A ClassCastException
if, after the type conversions above have been applied,
the resulting value is an object which is not an instance of a
subclass or subinterface of the erasure
(§4.6) of the corresponding formal
parameter type.
A NullPointerException
as a result of an unboxing conversion on a null
reference.
Neither strict nor loose invocation contexts include the implicit narrowing of integer constant expressions which is allowed in assignment contexts. The designers of the Java programming language felt that including these implicit narrowing conversions would add additional complexity to the rules of overload resolution (§15.12.2).
Thus, the program:
class Test { static int m(byte a, int b) { return a+b; } static int m(short a, short b) { return a-b; } public static void main(String[] args) { System.out.println(m(12, 2)); // compile-time error } }
causes a compile-time error because the integer
literals 12
and 2
have type
int
, so neither method m
matches under the rules
of overload resolution. A language that included implicit narrowing of
integer constant expressions would need additional rules to resolve
cases like this example.
String contexts apply only to an operand of the binary +
operator
which is not a String
when the other operand is a String
.
The target type in these contexts is always String
, and a string
conversion (§5.1.11) of the non-String
operand
always occurs. Evaluation of the +
operator then proceeds as
specified in §15.18.1.
Casting contexts allow the operand of a cast expression (§15.16) to be converted to the type explicitly named by the cast operator. Compared to assignment contexts and invocation contexts, casting contexts allow the use of more of the conversions defined in §5.1, and allow more combinations of those conversions.
If the expression is of a primitive type, then a casting context allows the use of one of the following:
If the expression is of a reference type, then a casting context allows the use of one of the following:
an identity conversion (§5.1.1)
a widening reference conversion (§5.1.5)
a widening reference conversion followed by an unboxing conversion
a widening reference conversion followed by an unboxing conversion, then followed by a widening primitive conversion
a narrowing reference conversion (§5.1.6)
a narrowing reference conversion followed by an unboxing conversion
an unboxing conversion (§5.1.8)
an unboxing conversion followed by a widening primitive conversion
If the expression has the null type, then the expression may be cast to any reference type.
If a casting context makes use of a narrowing reference conversion
that is checked or partially unchecked (§5.1.6.2,
§5.1.6.3), then a run time check will be performed
on the class of the expression's value, possibly causing a ClassCastException
.
Otherwise, no run time check is performed.
If an expression can be converted to a reference type by a casting conversion other than a narrowing reference conversion which is unchecked, we say the expression (or its value) is downcast compatible with the reference type.
The following tables enumerate which conversions are used in certain casting contexts. Each conversion is signified by a symbol:
≈ signifies identity conversion (§5.1.1)
ω signifies widening primitive conversion (§5.1.2)
η signifies narrowing primitive conversion (§5.1.3)
ωη signifies widening and narrowing primitive conversion (§5.1.4)
⇑ signifies widening reference conversion (§5.1.5)
⇓ signifies narrowing reference conversion (§5.1.6)
⊕ signifies boxing conversion (§5.1.7)
⊗ signifies unboxing conversion (§5.1.8)
In the tables, a comma between symbols indicates that a casting
context uses one conversion followed by another. The type Object
means any reference type other than the eight wrapper classes
Boolean
, Byte
, Short
, Character
, Integer
, Long
, Float
,
Double
.
Table 5.5-A. Casting to primitive types
To → | byte |
short |
char |
int |
long |
float |
double |
boolean |
---|---|---|---|---|---|---|---|---|
From ↓ | ||||||||
byte |
≈ | ω | ωη | ω | ω | ω | ω | - |
short |
η | ≈ | η | ω | ω | ω | ω | - |
char |
η | η | ≈ | ω | ω | ω | ω | - |
int |
η | η | η | ≈ | ω | ω | ω | - |
long |
η | η | η | η | ≈ | ω | ω | - |
float |
η | η | η | η | η | ≈ | ω | - |
double |
η | η | η | η | η | η | ≈ | - |
boolean |
- | - | - | - | - | - | - | ≈ |
Byte |
⊗ | ⊗,ω | - | ⊗,ω | ⊗,ω | ⊗,ω | ⊗,ω | - |
Short |
- | ⊗ | - | ⊗,ω | ⊗,ω | ⊗,ω | ⊗,ω | - |
Character |
- | - | ⊗ | ⊗,ω | ⊗,ω | ⊗,ω | ⊗,ω | - |
Integer |
- | - | - | ⊗ | ⊗,ω | ⊗,ω | ⊗,ω | - |
Long |
- | - | - | - | ⊗ | ⊗,ω | ⊗,ω | - |
Float |
- | - | - | - | - | ⊗ | ⊗,ω | - |
Double |
- | - | - | - | - | - | ⊗ | - |
Boolean |
- | - | - | - | - | - | - | ⊗ |
Object |
⇓,⊗ | ⇓,⊗ | ⇓,⊗ | ⇓,⊗ | ⇓,⊗ | ⇓,⊗ | ⇓,⊗ | ⇓,⊗ |
Table 5.5-B. Casting to reference types
To → | Byte |
Short |
Character |
Integer |
Long |
Float |
Double |
Boolean |
Object |
---|---|---|---|---|---|---|---|---|---|
From ↓ | |||||||||
byte |
⊕ | - | - | - | - | - | - | - | ⊕,⇑ |
short |
- | ⊕ | - | - | - | - | - | - | ⊕,⇑ |
char |
- | - | ⊕ | - | - | - | - | - | ⊕,⇑ |
int |
- | - | - | ⊕ | - | - | - | - | ⊕,⇑ |
long |
- | - | - | - | ⊕ | - | - | - | ⊕,⇑ |
float |
- | - | - | - | - | ⊕ | - | - | ⊕,⇑ |
double |
- | - | - | - | - | - | ⊕ | - | ⊕,⇑ |
boolean |
- | - | - | - | - | - | - | ⊕ | ⊕,⇑ |
Byte |
≈ | - | - | - | - | - | - | - | ⇑ |
Short |
- | ≈ | - | - | - | - | - | - | ⇑ |
Character |
- | - | ≈ | - | - | - | - | - | ⇑ |
Integer |
- | - | - | ≈ | - | - | - | - | ⇑ |
Long |
- | - | - | - | ≈ | - | - | - | ⇑ |
Float |
- | - | - | - | - | ≈ | - | - | ⇑ |
Double |
- | - | - | - | - | - | ≈ | - | ⇑ |
Boolean |
- | - | - | - | - | - | - | ≈ | ⇑ |
Object |
⇓ | ⇓ | ⇓ | ⇓ | ⇓ | ⇓ | ⇓ | ⇓ | ≈ |
Example 5.5-1. Casting for Reference Types
class Point { int x, y; } interface Colorable { void setColor(int color); } class ColoredPoint extends Point implements Colorable { int color; public void setColor(int color) { this.color = color; } } final class EndPoint extends Point {} class Test { public static void main(String[] args) { Point p = new Point(); ColoredPoint cp = new ColoredPoint(); Colorable c; // The following may cause errors at run time because // we cannot be sure they will succeed; this possibility // is suggested by the casts: cp = (ColoredPoint)p; // p might not reference an // object which is a ColoredPoint // or a subclass of ColoredPoint c = (Colorable)p; // p might not be Colorable // The following are incorrect at compile time because // they can never succeed as explained in the text: Long l = (Long)p; // compile-time error #1 EndPoint e = new EndPoint(); c = (Colorable)e; // compile-time error #2 } }
Here, the first compile-time error occurs because
the class types Long
and Point
are unrelated
(that is, they are not the same, and neither is a subclass of the
other), so a cast between them will always fail.
The second compile-time error occurs because a
variable of type EndPoint
can never reference a
value that implements the interface Colorable
. This
is because EndPoint
is a final
type, and a
variable of a final
type always holds a value of the same run-time
type as its compile-time type. Therefore, the run-time type of
variable e
must be exactly the
type EndPoint
, and type EndPoint
does not implement Colorable
.
Example 5.5-2. Casting for Array Types
class Point { int x, y; Point(int x, int y) { this.x = x; this.y = y; } public String toString() { return "("+x+","+y+")"; } } interface Colorable { void setColor(int color); } class ColoredPoint extends Point implements Colorable { int color; ColoredPoint(int x, int y, int color) { super(x, y); setColor(color); } public void setColor(int color) { this.color = color; } public String toString() { return super.toString() + "@" + color; } } class Test { public static void main(String[] args) { Point[] pa = new ColoredPoint[4]; pa[0] = new ColoredPoint(2, 2, 12); pa[1] = new ColoredPoint(4, 5, 24); ColoredPoint[] cpa = (ColoredPoint[])pa; System.out.print("cpa: {"); for (int i = 0; i < cpa.length; i++) System.out.print((i == 0 ? " " : ", ") + cpa[i]); System.out.println(" }"); } }
This program compiles without errors and produces the output:
cpa: { (2,2)@12, (4,5)@24, null, null }
Example 5.5-3. Casting Incompatible Types at Run Time
class Point { int x, y; } interface Colorable { void setColor(int color); } class ColoredPoint extends Point implements Colorable { int color; public void setColor(int color) { this.color = color; } } class Test { public static void main(String[] args) { Point[] pa = new Point[100]; // The following line will throw a ClassCastException: ColoredPoint[] cpa = (ColoredPoint[])pa; System.out.println(cpa[0]); int[] shortvec = new int[2]; Object o = shortvec; // The following line will throw a ClassCastException: Colorable c = (Colorable)o; c.setColor(0); } }
This program uses casts to compile, but it throws exceptions at run time, because the types are incompatible.
Numeric contexts apply to the operands of
arithmetic operators, array creation and access expressions,
conditional expressions, and the result expressions of switch
expressions.
An expression appears in a numeric arithmetic context if the expression is one of the following:
The operand of a unary plus operator +
, unary minus
operator -
, or bitwise complement operator ~
(§15.15.3, §15.15.4,
§15.15.5)
An operand of a multiplicative operator *
,
/
, or %
(§15.17)
An operand of an addition or subtraction operator for numeric types
+
or -
(§15.18.2)
An operand of a shift operator <<
, >>
, or >>>
(§15.19). Operands of these shift operators
are treated separately rather than as a group. A long
shift
distance (right operand) does not promote the value being
shifted (left operand) to long
.
An operand of a numerical comparison operator <
,
<=
, >
, or >=
(§15.20.1)
An operand of a numerical equality operator
==
or !=
(§15.21.1)
An operand of an integer bitwise operator &
, ^
, or |
(§15.22.1)
An expression appears in a numeric array context if the expression is one of the following:
An expression appears in a numeric choice context if the expression is one of the following:
Numeric promotion determines the promoted type of all the expressions in a numeric context. The promoted type is chosen such that each expression can be converted to the promoted type, and, in the case of an arithmetic operation, the operation is defined for values of the promoted type. The order of expressions in a numeric context is not significant for numeric promotion. The rules are as follows:
If any expression is of a reference type, it is subjected to unboxing conversion (§5.1.8).
Next, widening primitive conversion (§5.1.2) and narrowing primitive conversion (§5.1.3) are applied to some expressions, according to the following rules:
If any expression is of type double
, then the promoted
type is double
, and other expressions that are not of type
double
undergo widening primitive conversion to
double
.
Otherwise, if any expression is of type float
, then the
promoted type is float
, and other expressions that are not
of type float
undergo widening primitive conversion to
float
.
Otherwise, if any expression is of type long
, then the
promoted type is long
, and other expressions that are not
of type long
undergo widening primitive conversion to
long
.
Otherwise, none of the expressions are of type double
,
float
, or long
. In this case, the kind of context
determines how the promoted type is chosen.
In a numeric arithmetic context or a numeric array context,
the promoted type is int
, and any expressions that are not of
type int
undergo widening primitive conversion to int
.
In a numeric choice context, the following rules apply:
If any expression is of type int
and is not a constant
expression (§15.29), then the
promoted type is int
, and other expressions that are
not of type int
undergo widening primitive conversion
to int
.
Otherwise, if any expression is of type short
, and
every other expression is either of type short
or of
type byte
or a constant expression of type int
with
a value that is representable in the type short
, then
the promoted type is short
, and the byte
expressions
undergo widening primitive conversion to short
, and
the int
expressions undergo narrowing primitive
conversion to short
.
Otherwise, if any expression is of type byte
, and
every other expression is either of type byte
or a
constant expression of type int
with a value that is
representable in the type byte
, then the promoted type
is byte
, and the int
expressions undergo narrowing
primitive conversion to byte
.
Otherwise, if any expression is of type char
, and
every other expression is either of type char
or a
constant expression of type int
with a value that is
representable in the type char
, then the promoted type
is char
, and the int
expressions undergo narrowing
primitive conversion to char
.
Otherwise, the promoted type is int
, and all the
expressions that are not of type int
undergo widening
primitive conversion to int
.
Unary numeric promotion consists of applying numeric promotion to a single expression that occurs in a numeric arithmetic context or a numeric array context.
Binary numeric promotion consists of applying numeric promotion to a pair of expressions that occur in a numeric arithmetic context.
General numeric promotion consists of applying numeric promotion to all the expressions that occur in a numeric choice context.
Example 5.6-1. Unary Numeric Promotion
class Test { public static void main(String[] args) { byte b = 2; int[] a = new int[b]; // dimension expression promotion char c = '\u0001'; a[c] = 1; // index expression promotion a[0] = -c; // unary - promotion System.out.println("a: " + a[0] + "," + a[1]); b = -1; int i = ~b; // bitwise complement promotion System.out.println("~0x" + Integer.toHexString(b) + "==0x" + Integer.toHexString(i)); i = b << 4L; // shift promotion (left operand) System.out.println("0x" + Integer.toHexString(b) + "<<4L==0x" + Integer.toHexString(i)); } }
This program produces the output:
a: -1,1 ~0xffffffff==0x0 0xffffffff<<4L==0xfffffff0
Example 5.6-2. Binary Numeric Promotion
class Test { public static void main(String[] args) { int i = 0; float f = 1.0f; double d = 2.0; // First int*float is promoted to float*float, then // float==double is promoted to double==double: if (i * f == d) System.out.println("oops"); // A char&byte is promoted to int&int: byte b = 0x1f; char c = 'G'; int control = c & b; System.out.println(Integer.toHexString(control)); // Here int:float is promoted to float:float: f = (b==0) ? i : 4.0f; System.out.println(1.0/f); } }
This program produces the output:
7 0.25
The example converts the ASCII
character G
to the ASCII control-G (BEL), by
masking off all but the low 5 bits of the
character. The 7
is the numeric value of this
control character.